Server Testing

What is best service to use

[fashezee]

I would like to test our server speed. What is the best tool out there?

There are many online, some are super simple and others seem to flood you with info.
Is Google Insights Pagespeed sufficient?

[lammert]

The first thing you have to do is define your definition of speed. This can be multiple things like

• Time to first byte in browser
  
• Time until full loading is finished
  
• Time to fill the visible viewport on the screen

Each definition of speed may require different tuning of your Apache setup. When you defined what is important for your sites you can select a tool best for acquiring those metrics.

[robzilla]

Every tool you will find is flawed in some way. If you're serious about improving performance, you need to learn what affects load time, both at the frontend and the backend, so that you can gauge the advice given by the tools. Whatever you do, don't blindly follow that advice, and don't get hung up on their grading, trying to get a perfect score (it doesn't mean anything). That is not to say they can't give good advice, just that it's not always in the best interest of real-world performance.

Nevertheless, tools can give you a head start, and I would recommend PageSpeed Insights and Webpagetest.org. Avoid GTmetrix, its recommendations are outdated.

(Don't forget your browser has Developer Tools with a very helpful Network tab.)

If you're interested in reducing server response time (since you posted in the Apache forum), these tools won't be helpful and you'll need to look into your website's backend to increase the efficiency of your code (PHP, ASP, whatever you use), optimize database queries, implement caching where possible, and anything else you can think of.

[phranque]

Depending on your application you might also want to test your server under a significant load.

[fashezee]

Thank you all for the info.
GTmetrix does seem to provide an overload of info.

[tangor]

Tools are helpful, of course, but don't forget real world testing.

Ask your friends to hit the site via desktop, phone, whatever as each will provide a completely different experience which has nothing to do with your server speed ... unless your server just didn't serve it up in the first place.

Real world is where you what to be ... and knowing your target audience and what they are likely using at any given time.

There's "fast" and "just fast enough" and most of the time the latter makes more sense from a ROI point of view. At some point expense DOES matter.

[billzo]

If you are dynamically generating content using PHP or some other language, I would say that page generation time is the relevant metric. The way to check that is simple: just get the microtime at the top of your script, get it again at the end of the script, then subtract the end time from the start time to get the page generation time. Then put the page generation time in an HTML comment at the bottom of your page for your viewing pleasure.

Time to first byte is going to vary based on network traffic. That would probably be more of a relevant metric for static content like images, js and css files, etc., and certainly not for any dynamic content.

I have been using Pingdom Tools for years to check remote load times. Check it out.

[dougwilson]

Two things I've found with seo/speed/tests 1) there's no telling what the report says about them as a client - so three identical tests on three different services will likely give three different results 2) most dynamic sites use content from other servers/cdns (adsense, fonts, genericons...) and there's nothing we can do about that content, aside from remove it.

2.1) many webpage speed testers want you to pay them to optimize your site
2.2) ping your server, and yes, add something like [<?php printf(__('%d queries. %s seconds.', 'kubrick')] for wordpress

2.3) I guess WebmasterWorld could set up a place where we could test each others sites

[lucy24]

there's no telling what the report says about them as a client - so three identical tests on three different services will likely give three different results

Some sites offer an array of different starting locations. If you have no significant traffic from South Asia you obviously don't care about access time from Bangalore, but you do need to look at, say, both London and LA if you've got an English-language site. And then check at different times of day; it's reasonable for connections to be slower at 2PM than at 2AM (their time).